作者: Wang Xiaodan
DOI:
关键词:
摘要: The invention relates to a training method and system of commodity personalized ranking model. comprises the following steps: according long-term interest characteristics in historical data, carrying out off-line on model, obtaining parameter corresponding each characteristic, i.e. model with high precision, eliminating short-term data reduce time consumption; at unit interval, real-time expanding subjected training, on-line expanded obtain an updated characteristic characteristic. Therefore is for one interval higher timeliness realize balance precision so as better prediction result.